Good vs. Bad Prompts
What You’ve Learned in L2
Section titled “What You’ve Learned in L2”You’ve learned how a Prompt The input you send to an AI model — your question, instruction, or task. is structured, which techniques move you forward, how iteration works as a method, and how to tailor prompts to different task types. This lesson brings it all together — through five before-and-after comparisons.
Five Comparisons, Five Principles
Section titled “Five Comparisons, Five Principles”1. Vague vs. Specific
Section titled “1. Vague vs. Specific”| Weak Prompt | Strong Prompt |
|---|---|
| ”Write something about AI in healthcare." | "Write a 200-word summary of AI-assisted diagnostics for an audience of hospital administrators. Include one concrete example and one current limitation.” |
Why: A specific prompt defines the topic, length, audience, and content requirements. The AI doesn’t have to guess — it can deliver.
2. No Context vs. Full Context
Section titled “2. No Context vs. Full Context”| Weak Prompt | Strong Prompt |
|---|---|
| ”Give me feedback on this email." | "I’m writing to a client we’ve worked with for 3 years. They complained about a late delivery. Review my email for: professional but warm tone, clear apology, and a concrete resolution offer.” |
Why: Without context about the relationship, the situation, and the review criteria, feedback stays generic. With context, it becomes actionable.
3. Everything at Once vs. Focused Single Task
Section titled “3. Everything at Once vs. Focused Single Task”| Weak Prompt | Strong Prompt |
|---|---|
| ”Summarize the report, find insights, create an outline, and draft talking points." | "Read the report and identify the 3 most important takeaways for our sales team. Structure each takeaway as: key finding (1 sentence), supporting evidence (data from the report), recommended action.” |
Why: Four tasks at once produce four mediocre results. One focused task with a clear structure produces a strong output that you can build on iteratively.
4. No Format vs. Clear Output Format
Section titled “4. No Format vs. Clear Output Format”| Weak Prompt | Strong Prompt |
|---|---|
| ”Compare these project management tools." | "Create a comparison table for Asana, Monday, and Notion. Columns: Price (team plan), Integrations, Reporting, Learning Curve. One row per tool, ratings as bullet points.” |
Why: Without a format specification, the AI decides — often producing prose that’s hard to scan. With a table format, defined columns, and evaluation criteria, you get exactly the output you can use right away.
5. “Make It Better” vs. Targeted Revision
Section titled “5. “Make It Better” vs. Targeted Revision”| Weak Prompt | Strong Prompt |
|---|---|
| ”Improve this text." | "Revise this text: cut it by a third, replace passive constructions with active voice, and shift the tone from academic to ‘experienced practitioner briefing colleagues.’” |
Why: “Improve” is not an instruction — it’s a wish. Targeted iteration names what should change (length, style, tone) and gives the AI a clear direction.
L2 Summary
Section titled “L2 Summary”- Structure every prompt with role, task, and output format
- Use techniques like few-shot examples or chain-of-thought deliberately
- Refine results iteratively — with specific, directed feedback
- One task per prompt, one clear goal per task
- Match your prompt to the task type: analysis, creation, summary
- Give vague instructions without context, audience, or format specification
- Bundle multiple unrelated tasks into a single prompt
- Say 'Make it better' instead of steering with specific feedback
- Assume the AI knows your company, your goals, or your audience
- Use the same generic prompt regardless of the task
Your L2 Checklist
Section titled “Your L2 Checklist”Before moving on to L3, you should be able to answer each of these with “yes”:
- I can break a prompt into its building blocks (Role, Task, Context, Format)
- I know at least two prompting techniques and when to use them
- I iterate deliberately — with concrete feedback instead of vague change requests
- I tailor my prompts to the task type (analysis, writing, comparison, revision)
- I write instructions rather than questions when I need a specific output
What’s Next?
Section titled “What’s Next?”In L3 — Context as Infrastructure, you’ll learn how to move from individual conversations to persistent context. Instead of starting from scratch every time, you build knowledge that carries over across sessions.
The paradigm shift of L3: From single conversations to persistent context.